Every game

Fifty-seven rule changes, the most of any game in this study, with Scotus topping the table at +42. Four of five seats finished positive; only Ockham sat exactly at zero. A lightly-compressed table matching or beating its baseline on both volume and health is the study's first real surprise.

Fifty-two rule changes, the most of the two baseline games, with Rawls leading at +49 and four of five seats positive — only IbnKhaldun finished in the red, at -11. Two baseline runs this close together give the comparison's compression ladder a stable zero point to measure every altered rung against.

Thirty-nine rule changes and four of five seats finishing positive, led by Hume at +35 — solid, unremarkable, exactly what a baseline should be. The number that matters here isn't in this game alone: run under the current harness, this same table's baseline nearly doubled its adoption count compared to an older run on identical weights, a reminder that harness generation matters as much as the model underneath it.

Turing, Shannon, and vonNeumann each took exactly one turn — Babbage and Lovelace never got the chance — before the presiding judge produced an unenforceable ruling and the engine halted. Bookended by a full-length silent game and an early loop, this table's shortest outing is also its fastest collapse: three turns and the same turn-order trouble as its siblings.

The full turn limit ran out without a halt, but turn order broke down after the opening move: Shannon and vonNeumann traded every remaining turn between them, leaving Babbage and Lovelace silent for the entire fifty turns and Turing with only its opening move on the record. Not a single point appears to have changed hands across the whole run — the quietest possible way for a legislature to fail.

Twenty-five rule changes, and for the first time this table closed with anyone above zero: Pascal at +16 and Leibniz at +14, while Descartes, Berkeley, and Malebranche stayed negative. The gap between this size class and its smaller sibling looked wide after one game each; a third outing narrows it without closing it.

Nine rule changes — far fewer than this table's debut — and every seat closed negative, from Pascal's -48 down to Descartes' -100. A second data point for a table whose large end keeps producing legislation without producing a livable economy to go with it.

Twenty-six rule changes, nearly matching this table's most active outing, but still not a single seat above zero — AlFarabi again worst, at -136. Three games in, this table's small end has produced exactly one positive finisher — a solitary +5 in its debut — and nothing since; only the amount of lawmaking around that fact keeps changing.

Seven rule changes against twenty-five the first time out, and every seat finished negative — AlFarabi worst at -150, only AlGhazali holding at a flat zero. Volume and health both dropped together this time, reinforcing that this table's small end is erratic rather than reliably anything.

Thirty-one rule changes carried this table further than its first outing, and every seat closed positive — Dworkin and Nozick at the top on 73 and 72, Habermas trailing but still comfortably in the black at 48. The run ended the way its predecessor did: not a governance failure but an external request limit, this time arriving mid-verdict rather than between turns.

Identical seats to this table's collapse game, but with a steadier judge in the chair: adoptions rose from one to six and the floor improved from -180 to -84. The economy still failed completely — nobody finished positive either time — so the experiment cleanly separates two failures that had looked like one: the judge was genuinely chaotic and swappable, but this table's own legislature is weak on its own terms, judge aside.

Frege topped the table again at +47, with Quine and Wittgenstein also finishing positive; Carnap closed flat at zero and only Russell, at -22, fell into the red. Forty-two adopted rules against a first outing's forty-nine — both prolific, both mostly healthy. Two outings, nearly the same shape: this table's whole signature travels well.

Twelve rule changes and a floor of -130 for Locke, with Rousseau close behind at -90 — a different failure shape from this table's first try, which simply flatlined at zero. Across its outings this table's signature is holding: it does not build a positive economy, but exactly how it fails varies game to game.

A full fifty turns and sixteen adopted rules produced total economic inertia — Hobbes, Locke, Rousseau, Kant, and Wollstonecraft all closed exactly where they started. Legislating steadily without ever actually moving anyone's score is its own particular failure mode, distinct from the deep-negative collapse this table has produced elsewhere.

Twenty-three rule changes, and this time Voltaire — the same seat that closed one outing at +881 — finished at -157, the table's worst score. Only Montaigne and Condorcet stayed in single-digit positive territory. Three outings for this table now read as high variance and stark inequality rather than one characteristic shape; who ends up on top or bottom is not predictable game to game.

Twelve rule changes produced the opposite shape from this table's highest-profile outing: Diderot led at +79 and Voltaire followed at +76, with only Montaigne, at -16, finishing negative. A ninety-five-point gap between top and bottom is still wide, but it is a spread, not a runaway.

Nine rule changes, three of five seats positive again — but this time HanFeizi topped the table at +30, not Mencius, who finished flat at zero. Two straight wins for the same seat had looked like a signal; a third game showed it was coincidence. What holds across all three outings is the table's economy, not any one seat's dominance.

Six rule changes this time against thirteen the first time out, and the economy held up in miniature: three of five seats finished positive, led by Mencius at +40. Zhuangzi's -10 was the only real sore spot. A second data point for a table whose reputation is 'reliably positive,' not 'reliably record-setting.'

In eleven turns this table amended seven of the ruleset's foundational articles — defeated-proposal semantics, the adoption threshold, turn structure, negotiation, disputes, the win condition, and voting itself — a constitution-wide overhaul rather than the single-rule tinkering smaller tables favor. All five seats were still positive and close together, 9 to 12 points, when a request quota, not a governance failure, ended the run. A genuinely different signature cut short before it could show its ceiling.

The collapse table: a single adopted rule change in two hundred turns, three seats driven to -650 or worse, and two more sitting at a flat zero for want of anything to do. This is the floor the archive's size comparisons use as their worst case — a legislature that essentially never got going.